NOVEMBER 2015 1

Performance limitations of 40 Gb/s SSB MB-OFDMmetropolitan networks induced by phase-to-intensity

conversion of laser phase noiseMiguel Sousa Pereira No69044

Instituto Superior TecnicoAv. Rovisco Pais, 1049-001 Lisboa, Portugal

E-mail: miguelpereira@tecnico.ulisboa.pt

Abstract—Multi-band (MB) orthogonal frequencydivision multiplexing (OFDM) signals have recentlybeen proposed to be used in optical metropolitan(metro) networks. MB-OFDM signals transmissionprovides increased bandwidth allocation flexibility,high spectral efficiency, higher capacity provisioninggranularity and high tolerance to linear fibre distor-tion effects. Metro network implementation cost re-striction can be meet by employing direct-detection(DD) OFDM. However, although DD-OFDM presentsrobustness against phase fluctuations, fibre chromaticdispersion (CD) causes phase-to-intensity conversion oflaser phase noise, leading to performance degradationin DD-OFDM systems.

The objective of this work is to evaluate throughnumerical simulation the maximum laser linewidth andmaximum reach of a 40 Gb/s single sideband (SSB)MB-OFDM metro ring limited by laser phase noise.

In this work, the transmission of MB-OFDM signalsemploying virtual carriers is studied. For a 3-band MB-OFDM system with laser linewidths as high as 5 MHz,it is demonstrated that the penalty of the requiredoptical signal-to-noise ratio (OSNR) to achieve a biterror rate (BER) of 10−3 after a fibre span of 500 kmdoes not exceeds 1 dB. The results considering a realnetwork show that, using a Gaussian BS, it is possible totransverse 5 fibre spans of 40 km with a maximum laserlinewidth of 10 kHz for BER < 10−3. A highly selective2nd-order super Gaussian BS is also evaluated and it isshown that the MB-OFDM signal can travel throughmore than 7 spans of 40 km for laser linewidths up to100 kHz.

Index Terms—Metro networks, orthogonal frequencydivision multiplexing, multi-band, direct detection,laser phase noise, phase-to-intensity noise conversion.

I. Introduction

USUALLY, telecommunication networks are strati-fied in core, metro and access networks. The core

or backbone network deals with long-haul transmissions,covering vast distances. It provides connectivity to metronetworks and carries large traffic. Access networks arethe connection to end-users, support a wide diversity ofprotocols and cover the last 10 km to 35 km of the overallnetwork.

Metro networks are responsible for delivering trafficfrom core to access networks and vice-versa, but also

between different access networks. Designed to serve largeand densely populated areas, metro networks typicallyextend between 200 km and 300 km, depending on theregion geography [1]–[3].

The metro network must aggregate different types oftraffic coming from the access layer. Therefore, flexibility,scalability, dynamic reconfiguration and transparency arekey features [4]. Also, since the metro network infrastruc-ture is shared among fewer people than core networks, itrequires cost-effective solutions [4].

The most common metro networks present a ring topol-ogy [1], [5], as shown in figure 1. An optical network iscomposed by nodes linked by optical fibre, generally thestandard single mode fibre (SSMF) [2], [8]. The nodesconnect the metro network to the access or core networksand typically the distance between two adjacent nodes (aspan) ranges from 5 km to 100 km [2], [9].

Fig. 1: Metro network with five nodes in a typical ringconfiguration.

Since bandwidth demand continues to grow exponen-tially, there is little alternative but to develop new so-lutions that allow us to maximize existing systems andnetworks. Subsequently, wavelength division multiplexing(WDM) role is increasingly significant as metro networkstend to adopt optical transparency. As dense WDM(DWDM) is deployed extensively in metro networks, theuse of a large number of wavelengths and unpredictablebandwidth demand have led to the development of recon-figurable OADMs (ROADMs), allowing wavelength rear-rangement through software control [10].

Moving towards all or predominantly optical environ-ments enables several advantages. The most significant ad-

2 NOVEMBER 2015

vantage is the revenue potential that comes with managingwavelengths instead of fibre cable-laying. Nevertheless, intransparent networks, access to tighter granularity levelsthan the wavelength is still performed resorting to an elec-trical layer. Otherwise, nodes would be unable to manage,process and adjust traffic for optimization of the systemcapacity. Therefore, future metro networks aim at opticalrouting with granularity at the sub-wavelength level [4].

MB-OFDM considers the transmission of multiple nar-row and independent OFDM bands in each wavelength.This implementation presents many advantages. Thesharp shape of the OFDM spectrum allows achieving highspectral efficiency by reducing the spacing between bands[4]. With the use of parallel lower-speed bands, electricalbandwidth requirements can be reduced [11]. Furthermore,employing DD makes MB-OFDM a very cost-effectivesolution for metro networks.

High spectral efficiency and resilience against linear fibreeffects led OFDM systems to be widely implemented in op-tical communications [13]–[18]. Interest in optical OFDMis also supported by recent advances in microelectronictechnologies, such as the analogue-to-digital converter(ADC), digital-to-analogue converter (DAC) and digitalsignal processing (DSP) [18].

OFDM systems are composed by radio frequency (RF)transmitters and receivers. Thus, to accomplish opticaltransmission, the OFDM transmitter is followed by aelectrical-optical (EO) converter and the OFDM receiver ispreceded by a optical-electrical (OE) converter. Dependingon how the OE conversion is performed the system isclassified as either coherent optical OFDM (CO-OFDM),when the received signal is mixed with a locally gener-ated optical carrier, or as DD-OFDM, when the signal istransmitted with the optical carrier [13]–[18].

A DD-OFDM implementation uses a simpler receiverwith a single photodetector and without a local laser, thuscheaper. However, it requires more optical power in orderto transmit the optical carrier and it has less spectralefficiency since some optical frequencies are left unused asa guard band between the optical carrier and the OFDMsubcarriers to avoid mixing products interference. Nev-ertheless, DD-OFDM is preferable for metro and accessnetworks where cost is the primary concern [13], [18], [19].

Recently, a 100 Gb/s DD MB-OFDM superchannelsystem with virtual carriers was proposed for long-haulnetworks [20]. The system described in [20] uses dualcarriers at both superchannel sides to assist DD. ThisMB-OFDM system is quite challenging to implement inflexible metro networks due to the huge receiver front-end bandwidth requirements and the need for a dual-band optical filter to assist detection. One way to avoidusing a dual-band optical filter and simultaneously reducethe required receiver bandwidth is to use a single virtualcarrier close to each OFDM band to assist the banddetection [4], [21], [22].

The MB-OFDM system considered in this work is sup-ported in the recently proposed metro networks basedon SSB MB-OFDM signals (MORFEUS) [4], [21], which

employ virtual carriers to assist DD. MORFEUS proposaladdresses the future requirements of metro networks suchas high flexibility, scalability, dynamic reconfigurabilityand transparency.

The MORFEUS solution presents some challenges. Thecapacity and number of bands of the MB-OFDM signal isdetermined by the cost restriction of electrical componentsin metro networks. The increased complexity caused byband insertion and extraction blocks and their impact onthe system performance still needs to be investigated. Thereduced selectivity of the optical filters used in extractionleads to inter-band crosstalk. Inefficient band blockingresults in intra-band crosstalk between an extracted bandthe newly inserted band.

Additionally, DD also presents some drawbacks. Partic-ularly, significant performance limitations are imposed byphase noise and wavelength drift of laser sources. AlthoughDD systems are robust to phase impairments, due tofibre CD, phase noise is converted into intensity noisedegrading the MB-OFDM signal. While phase noise effectshave been widely studied in CO-OFDM systems [23], [24],[26]–[29], the impact of laser phase noise on the DD-OFDM system performance has only been analysed forlong-haul networks transmitting a single band [30]. Thus,it is quite important to evaluate the performance of metronetworks employing DD limited by the finite linewidthof laser source (which results mainly from phase noise),particularly when transmitting SSB MB-OFDM signals.

The main objectives of this work are to assess (i) thedegradation induced by the conversion of the laser phasenoise to intensity noise realized by the fibre dispersion in40 Gb/s SSB MB-OFDM metro networks and (ii) the max-imum laser linewidth and maximum reach of the 40 Gb/sDD MB-OFDM metro ring limited by laser phase noise.In order to achieve the proposed objectives, a numericalsimulator was developed in MATLAB.

The rest of this paper is organized as follows. In sectionII, the MB-OFDM system model is provided. SectionIII describes the laser phase noise and its effects. Insection IV, numerical results and analysis of the systemperformance in the presence of laser phase noise are given.Finally, section V concludes this paper.

II. MB-OFDM system model

The classical multi-carrier modulation (MCM) system,e. g. frequency-division multiplexing (FDM), is based onnon-overlapped band-limited signals. A MCM system canbe implemented using a large number of oscillators and fil-ters at both transmitter and receiver which, in turn, leadsto channel spacing becoming multiple times the symbolrate, greatly reducing the signal spectral efficiency [31].Alternatively, OFDM signals use overlapped subcarriersto increase spectral efficiency but in an orthogonal way,with each subcarrier orthogonal to every other subcarrier,and so preventing inter-carrier interference (ICI).

PEREIRA: LIMITATIONS OF MB-OFDM METRO NETWORKS BY PHASE-TO-INTENSITY CONVERSION OF LASER PHASE NOISE 3

A. The OFDM signalConsidering that a OFDM transmitted signal has a set

of subcarriers, each one bearing information symbols, thenthat signal can be expressed as a time-domain sum of allsubcarriers [31], given by

s(t) =+∞∑i=−∞

Nsc∑k=1

ckisk(t− iTs) (1)

where cki is the information symbol at the k-th subcarrierof the i-th OFDM symbol, sk(t) is the waveform of thek-th subcarrier, Nsc is the number of subcarriers and Tsis the OFDM symbol duration.

A typically used function set is the windowed discretetones [31], given by

sk(t) = Π(t)ej2πfkt (2)

where fk is the frequency of the k-th subcarrier and Π(t)is the pulse shaping function given by

Π(t) ={

1, if 0 < t ≤ Ts0, if t ≤ 0, t > Ts .

(3)

The OFDM signal’s orthogonality can be verified if thefollowing condition [31] is fulfilled

fk − fl = m

Ts, m ∈ N . (4)

In OFDM, a large number of subcarriers (Nsc) is re-quired in order to the transmission channel appears flatto each subcarrier and therefore recover the subcarrierswith minimum signal processing complexity [18]. Never-theless, a large number of subcarriers leads to an extremelycomplex architecture involving many oscillators and filters.Another way to generate the orthogonal subcarriers isto make use of the discrete Fourier transform (DFT),applying the inverse DFT (IDFT) to implement OFDMmodulation and DFT for demodulation. Focusing on onlyone OFDM symbol, (i = 0), and assuming a samplingperiod of Ts

Nsc, then, from equation (1), the m-th sample

of the transmitted signal becomes

sm =Nsc∑k=1

ck · ej2πfkmTsNsc , m = 0, 1, ..., Nsc − 1 . (5)

Using the orthogonality condition presented in equation(4) and the convention [18] that the frequency of the k-thsubcarrier is given by

fk = k − 1Ts

(6)

then, equation (5) can be rewritten as

sm =Nsc∑k=1

ck · ej2π(k−1)m

Nsc = IDFT{ck} (7)

where IDFT{x} stands for the IDFT of x.

B. OFDM transmission systemThe structure of a generic OFDM communication sys-

tem based on a DFT architecture is depicted in figure2. Both the transmitter and receiver process the signaldigitally, while transmission is done in the analogicaldomain. The fast Fourier transform (FFT) and inverseFFT (IFFT) are used as practical versions of the DFT andIDFT, respectively. At the transmitter side, binary data ismapped into symbols accordingly to a digital modulationscheme. In this case, data is mapped using QAM.

Fig. 2: Block diagram of DFT-based OFDM transmitterand receiver.

The resulting sequence of QAM symbols must be con-verted to a suitable input for the IFFT. This operationis performed in the serial-to-parallel (S/P) block, wherea sequential input is divided in multiple Nsc parallelstreams. The IFFT modulates the mapped symbols intosubcarriers of a OFDM symbol. After the Nsc-point IFFT,a cyclic prefix (CP) is added to the beginning of theOFDM symbol [32].

The OFDM symbol is lined up through a parallel-to-serial (P/S) process. DSP facilitates the OFDM implemen-tation but requires digital-analogue conversion. Therefore,the signal must be converted to analogue by a DACwhich turns discrete samples used in the digital domaininto a continuous analogue waveform. Also, the signal isfiltered by a low-pass filter (LPF) to eliminate excessivebandwidth, due to aliasing [31].

The resulting signal is up-converted by an in-phase/quadrature (I/Q) modulator. Taking advantage ofthe real (in-phase) and imaginary (quadrature) signalcomponents, the modulator up-converts the signal froma baseband frequency into a passband one used for trans-mission.

At the receiver end, the reverse process must be donein order to obtain the original data. Firstly, the analoguereceived signal is down-converted by an I/Q demodulator.The next step is a digital conversion using a ADC enabling

4 NOVEMBER 2015

DSP. The ADC samples a received waveform in such waythat with an ideal channel the ADC output should beequal to the DAC input.

The digital OFDM signal received is separated in Nsc+CP streams by a S/P block and the CP is removed.We now have a suitable input for the Nsc-point FFT.The FFT demodulates the OFDM subcarriers into QAMsymbols. The received QAM symbols must be equalizedand afterwards those symbols are demapped through adecision process where the original binary data is expectedat the output of the demapper.

C. SSB MB-OFDM transmission systemIn metro networks using DD, after photodetection, the

fibre CD accumulation induces high power fading in doublesideband (DSB) signals [33]. A possible solution to over-come CD-induced power fading is to transmit SSB signals[4] at the expense of reducing the available bandwidthof the OFDM signal. Using SSB signals, that can begenerated at the optical transmitter, instead of resorting tooptical dispersion compensation allows network upgradingwithout modifying equipments installed between transmit-ter and receiver.

SSB OFDM transmitters generally reserve a frequencygap between the optical carrier and the OFDM signal.The gap bandwidth is usually similar to the OFDM signalbandwidth [17], [20], [22], [35]–[37]. This gap serves toaccommodate the distortion induced by signal-signal beatinterference (SSBI) resulting from the DD process, asdetailed in subsection II-C3. However, this leads to aspectrally inefficient system.

The inclusion of a virtual carrier (RF tone) close to eachOFDM band allows reducing the frequency gap and, thus,increasing the spectral efficiency. The main limitation ofthis technique is the additional distortion of the OFDMsignal caused by the SSBI. Nevertheless, such distortioncan be diminished either by increasing the power of thevirtual carrier compared to the power of its correspondingOFDM band or by using DSP algorithms at the receiverside to reconstruct and remove the SSBI term from thephoto-detected signal [8], [38], [39].

Figure 3 presents a scheme of the optical SSB MB-OFDM system employing virtual carriers (VCs). As shownin figure 3, the MB-OFDM transmitter is composed byOFDM transmitters, one for each band, where virtualcarriers are electrically generated and added to the OFDMbands. Additionally, the MB-OFDM transmitter is re-sponsible to convert the generated signal into an opticalSSB signal. The MB-OFDM signal is transmitted alonga SSMF span and the necessary signal amplification isassured by erbium-doped fibre amplifiers (EDFAs). Atthe MB-OFDM receiver, the desired band is filtered fromthe MB signal by a band selector (BS). The MB-OFDMreceiver also includes a DD-based OE converter, a digitalSSBI removal block and a OFDM receiver.

1) Virtual carrier-assisted transmission: The virtualcarrier enables to separately detected each OFDM band

Fig. 3: Scheme of the optical SSB MB-OFDM systememploying DD and virtual carriers to assist detection.

[21]. The BS in the optical receiver selects both the OFDMband and the virtual carrier and, after photo-detection, theresulting direct current (DC) component is removed fromthe signal.

Figure 4 illustrates the spectral occupancy of a (SSB)MB-OFDM signal with virtual carriers in the electricaldomain. In figure 4, BOFDM is the bandwidth of a OFDMband considering that all bands have the same bandwidth,fc,n and fvc,n are the central frequency and the frequencyof the virtual carrier, respectively, of the n-th OFDMband. Additionally, figure 4 identifies the virtual carrier-to-band gap (VBG), the band gap (BG) and band spacing,∆fc, which is the spacing between central frequencies ofadjacent bands.

Fig. 4: Spectral scheme of a 3-band MB-OFDM signalemploying virtual carriers in the electrical domain.

The VBG is defined as the width of the frequency gapbetween the OFDM band and its corresponding virtualcarrier. The VBG should be as small as possible in orderto maximize spectral efficiency but, since it coincides withthe SSBI accommodation gap, if the VBG is narrowerthan the bandwidth of the OFDM band then, after photo-detection, SSBI overlaps the signal. Assuming that theMB-OFDM receiver completely removes SSBI after pho-todetection, then the VBG is selected in order to allocatethe virtual carrier at a null of the OFDM subcarriers sinc-square spectrum avoiding the out of band side lobes and

PEREIRA: LIMITATIONS OF MB-OFDM METRO NETWORKS BY PHASE-TO-INTENSITY CONVERSION OF LASER PHASE NOISE 5

unnecessary degradation. Thus, the VBG can be obtainedas follows

VBG = kRb

Nband

1Nsc log2 M

, k ∈ N . (8)

The BG is defined as the width of the frequency gapbetween two adjacent bands of the MB-OFDM signal. Anappropriate BG is necessary to avoid inter-band crosstalkdue to the finite selectivity of the BS. The BG value isobtained considering that: (i) the maximum bandwidthfor the MB-OFDM signal is 20 GHz [4] and (ii) the bandspacing is chosen targeting an even distribution of theOFDM bands along the MB-OFDM signal spectrum. TheBG relation to the band spacing is perceived in figure 4and it is given by

∆fc = BG +BOFDM . (9)The virtual carrier amplitude depends on the virtual

carrier-to-band power ratio (VBPR), which is defined asthe ratio between the power of the virtual carrier, pvc, andthe power of the OFDM band, pband. The VBPR expressedin dB is given by

VBPR = 10 log10

(pvcpband

). (10)

2) Optical transmitter: An optical transmitter is re-quired to convert the electrical OFDM signal into opticalform and launch the resulting optical signal into theoptical fibre. In this work, an optical transmitter usingexternal modulation is employed.

The optical SSB signal can be generated by using aconventional optical modulator followed by an optical SSBfilter to remove one sideband of the signal or using adual parallel (DP) Mach-Zehnder modulator (MZM) [40],avoiding the use of an optical SSB filter which presents alimitation for spectral efficiency and system performancedue to its finite selectivity. The DP-MZM EO converter iscomposed by two inner MZMs and an outer MZM.

The DP-MZM allows generating a SSB version of theMB-OFDM signal by applying the electrical signal and itsHilbert transform to the two arms of the DP-MZM [41].The ideal Hilbert transform transfer function is given byHH(f) = j ·sgn(f). In a real system, the Hilbert transformis implemented by a hybrid coupler.

Figure 5 shows the spectrum of the optical SSB MB-OFDM signal without optical carrier at the output ofthe DP-MZM. The plot shows the aforementioned virtualcarriers close to each OFDM band. Due to the non-idealHilbert transform implemented by the hybrid coupler, thesignal at the output of the DP-MZM is not a pure SSBsignal as vestiges of the the lower sideband signal can stillbe found. Nevertheless, the lower sideband is significantlyattenuated, about 15 dB, in comparison with the uppersideband.

3) Optical receiver: The optical receiver OE conversionis done by a photo-detector. Usually, in optical fibrecommunication systems, the photo-detectors are semicon-ductors photo-diodes [43]. In this work, the optical receiveruses a positive-intrinsic-negative (PIN) photo-diode.

−20 −15 −10 −5 0 5 10 15 200

10

20

30

40

50

60

70

80

90

ν − ν0 [GHz]

Norm

alizedPSD

[dB]

Fig. 5: Spectrum of the optical MB-OFDM signal at theDP-MZM output.

A single OFDM band signal, in a back-to-back imple-mentation, without BS, without any noise addition andneglecting non-linearities of the EO conversion, presents aphoto-current at the PIN output given by

iPIN(t) = A2vc + 2Avc<{sOFDM(t)}+ |sOFDM(t)|2 (11)

where Avc is the virtual carrier amplitude and sOFDM(t)is the OFDM band signal, both after the OE conversion.In equation (11), the first term is a DC component (easilyremoved with a DC block), the second term is the desiredOFDM band signal amplified by a constant value and thethird term is the SSBI.

In the developed simulator, SSBI removal is achieved byobtaining the SSBI-induced distortion separately and thenremoving it from the photo-current signal, as proposed in[42].

III. Laser phase noise

In optical systems, phase noise is caused by fluctuationin the laser source, in this case a continuous wave (CW)laser. An ideal optical source without any modulationis a monochromatic oscillator. Its output electrical fieldpower spectral density (PSD) is a line located at theoscillation frequency and its spectral linewidth is null [43].Although photon generation in a laser is done throughstimulated emission, in a non-ideal situation, there is somespontaneous emission, causing phase fluctuations in lasers.Consequently, the optical source is noisy and can be seenas not monochromatic or incoherent [43]. This means that,even in absence of modulation, the output field PSD isnot a line and no longer has null spectral linewidth. Insemiconductor lasers, the linewidth is usually large, up tothe megahertz range [31].

The phase fluctuation of a laser, φ(t), which can bequantified by the laser linewidth, ∆νL, is generally mod-

6 NOVEMBER 2015

elled by the Wiener process [23]–[28] as follows

φ(t) = 2π∫ t

0n(v)dv (12)

where n(t) is a zero mean random process, known as whiteGaussian noise, with a variance of ∆νL/(2π) [30]. As aresult, the phase fluctuation has a normal distributionwith zero mean and a variance of ∆νLt/(2π) [30].

In a DD-OFDM system, the transmitter’s oscillator isthe only phase noise source. Furthermore, due to thepower-detection nature of the photo-diode, the laser phasenoise will not appear in the converted electrical signal andthus it is harmless to the system performance. However,in optical transmission, the signal accumulates a certainamount of fibre CD and its different subcarriers experiencedifferent timing offset [30].

After photo-detection the received signal can be ex-pressed, resorting to equation (11), as follows [30]

iPIN(t) = DC+2Avc<{sOFDM(t)ejρk(t)ejθk}+SSBI . (13)

At the photo-diode output, the desired signal comesmultiplied by the converted phase noise (CPN) term,ejρk(t), where the converted phase fluctuation on the k-th subcarrier, ρk(t), results from the combination of thelaser phase noise and the fibre CD time delay [30]. Thedesired signal comes also multiplied by ejθk , where θk isthe CD-induced phase rotation on the k-th subcarrier [30].

The subcarriers symbols of the photo-detected signalwith phase noise are not only corrupted by the phaserotation term (PRT), ejρk(t)ejθk , but also by informationof the adjacent subcarriers [30]. The latter effect has theappearance of a zero mean Gaussian noise [23], [24], [26],[30], normally known as ICI or loss of orthogonality [27],[29], [30].

Additionally, since signal power is leaking to the PRTand ICI, the presence of phase noise is responsible forpower degradation in the received subcarriers.

In coherent-detection systems, symbols phase rotationinduced by phase noise is known as common phase error(CPE) [23] and it can be corrected by phase rotationor equalization. The CPN is dependent of the subcarrierfrequency and its bandwidth is both independent of thelaser linewidth and a function of the time delay [30]. Thus,the CPN bandwidth might range in orders greater than100 MHz [30] and such a broad spectrum introduces sig-nificant ICI. Particularly, PRT differs from CPE becauseit varies with the subcarrier frequency and therefore is notcommon to all subcarriers. PRT power is found to increasewith fibre length and decrease with the total number ofsubcarriers, but only affecting the imaginary part of thesignal and therefore resulting in a phase rotation [30].

Figure 6 illustrates the described effects of the phasenoise on the received QAM symbols. The received QAMsymbol, shown in figure 6, suffers from power degradation,thus its vector is shorter than the ideal QAM symbol.PRT effect is represented by a dashed curve indicating thesymbol rotation. In figure 6, the QAM symbol scattering

due to ICI is illustrated by a circle of possible locationsfor the received symbol.

Fig. 6: Phase noise effects on the received QAM symbol,adapted from [30].

Figure 7 shows the received QAM symbols of one OFDMband after a 280 km fibre span, with and without phasenoise. The numerical simulator parameters considered forthe single OFDM band system are presented in table I.

Tab. I: Parameters of the single OFDM band system.Bit rate [Gb/s] 14.3

Bandwidth [GHz] 3.6Number of subcarriers 128

QAM mapping 16-QAMVBG [MHz] 27.9VBPR [dB] 6

The selected fibre length corresponds to about thetypical maximum length of a metro ring, e. g., a seven-node ring with fibre spans of 40 km between nodes.

The constellation presented in figure 7a) shows thefibre CD-induced rotation. Figure 7b) shows both theconstellation rotation and the ICI noisy-like effect of phasenoise, now visible due to the combination of phase noiseand fibre CD.

IV. Performance analysisIn addition to the parameters presented in table I, the

performance analysis of the MB-OFDM system considersthe parameters presented in table II.

Two different approaches are considered to evaluatethe MB-OFDM signal performance degradation due tothe combination of phase noise and fibre CD. The firstapproach consists on a noise loading circuit, where theOSNR is enforced to the system. In this approach theamplified spontaneous emission (ASE) noise is added to

Tab. II: Parameters of the MB-OFDM system.Number of OFDM bands 3

Total bit rate [Gb/s] 42.8Band spacing [GHz] 6.25

BS Gaussian 2nd-order super GaussianBS bandwidth [GHz] 3 4

fc,1 [GHz] 3.5

PEREIRA: LIMITATIONS OF MB-OFDM METRO NETWORKS BY PHASE-TO-INTENSITY CONVERSION OF LASER PHASE NOISE 7

−5 −4 −3 −2 −1 0 1 2 3 4 5−5

−4

−3

−2

−1

0

1

2

3

4

5

In-phase component

Quadratu

recomponent

(a)

−5 −4 −3 −2 −1 0 1 2 3 4 5−5

−4

−3

−2

−1

0

1

2

3

4

5

In-phase component

Quadratu

recomponent

(b)

Fig. 7: Constellations of the received 16-QAM symbolsafter a 280 km fibre span, without BS and equalizer,with (a) null and (b) 5 MHz laser linewidth. Opticalamplification and electrical noises are not considered.

the signal at the BS input and the performance limita-tions introduced by the network intermediate nodes areneglected (single span). The second approach considersa real network, where the OSNR is not predefined anddepends on noise accumulation along the network ring.Contrarily to the noise loading circuit implementation,this approach considers that noise is distributed along thenetwork, at each EDFA the signal accumulates more ASEnoise and amplifies both the signal and already existingnoise.

The purpose of using a noise loading circuit is to assessthe required OSNR value to obtain the target BER of10−3. Figure 8 shows, for different laser linewidths, thevariation relative to back-to-back (B2B) of the requiredOSNR (∆OSNR) as function of the SSMF length.

Although, the maximum length of a metro networkis typically around 300 km, figures 8a) and 8b) presentresults up to 500 km in order to observe the phase-to-intensity noise conversion effect due to the combination oflaser phase noise and fibre CD. Nevertheless, the SSMFlength by itself does not seem to affect much the systemperformance and its effect is more significant for laserlinewidths above 1 MHz.

Figure 8a) shows that, when employing a Gaussian BS,

0 100 200 300 400 500

0

0.2

0.4

0.6

0.8

1

SSMF length [km]

∆OSNR

[dB]

0 MHz

3 MHz

5 MHz

(a) Gaussian BS

0 100 200 300 400 500

0

0.1

0.2

0.3

0.4

0.5

SSMF length [km]

∆OSNR

[dB]

0 MHz

3 MHz

5 MHz

(b) 2nd-order super Gaussian BS

Fig. 8: Variation of the required OSNR to achieve a BERof 10−3 as a function of the SSMF length for different laserlinewidths, employing (a) a Gaussian filter and (b) a 2nd-order super Gaussian filter as BSs.

for a laser linewidth of 5 MHz, the ∆OSNR only reaches1 dB at 500 km of SSMF. Figure 8b) shows that, using a2nd-order Gaussian BS, the ∆OSNR for each linewidth isreduced to less than a half in comparison with figure 8a).These results show that, when phase noise is considered,fibre CD is quite less significant to the system performancethan the BS. The behaviour of the OSNR penalty variationwith the fibre length, shown in figure 8, is in agreementwith the study realized in [30].

In order to support the previous results, figure 9 presentsthe error vector magnitude (EVM) [44] variation along thesubcarriers of the OFDM band with worst performance,when employing a Gaussian BS.

In figure 9a) the first and last subcarriers have higherEVM values. This is due to the BS shape and narrowbandwidth, which leads to a EVM variation from -40 dBup to -30 dB, even in a B2B situation without phasenoise. Figure 9b) shows that after a 500 km span, fibreCD contributes with an evenly distributed EVM increaseacross the subcarriers to values between -35 dB and -25dB. Thus, fibre CD by itself affects less the OFDM bandperformance than the selectivity of the BS.

Figure 9c) shows the combined effect of the BS and thelaser phase noise. In comparison with figure 9a) the EVMacross the subcarriers greatly increases, with a maximumdifference of approximately 15 dB. This is due to phase-to-intensity conversion of laser phase noise induced by the

8 NOVEMBER 2015

0 50 100 128−45

−40

−35

−30

−25

−20

−15

Subcarrier

EVM

[dB]

(a)

0 50 100 128−45

−40

−35

−30

−25

−20

−15

Subcarrier

EVM

[dB]

(b)

0 50 100 128−45

−40

−35

−30

−25

−20

−15

Subcarrier

EVM

[dB]

(c)

0 50 100 128−45

−40

−35

−30

−25

−20

−15

Subcarrier

EVM

[dB]

(d)

Fig. 9: EVM in dB as a function of the subcarriers (a) inB2B, (b) after a fibre span of 500 km, (c) in B2B and with5 MHz of laser linewidth, (d) after a fibre span of 500 kmand with 5 MHz of laser linewidth. A Gaussian filter BSis employed.

BS. Since the aim of this work is mainly to analyse fibreCD-induced phase-to-intensity noise conversion of laserphase noise, further analysis of the BS in the presenceof phase noise is out-of-scope of this paper and shouldappear in other publication. The situation shown in figure9d) considers the combination of phase noise, a BS andfibre CD. The EVM increases for all subcarriers, reachinga maximum increase of 17 dB in comparison with the situ-ation shown in figure 9a). Comparing with figure 9c) case,fibre CD only contributes with a more even degradationacross the subcarriers.

The following simulation results consider the secondperformance evaluation approach, where the MB-OFDMsignal is transmitted between network nodes along SSMFspans of 40 km.

In order to assess the performance differences betweenthe BS solutions, figure 10 presents the EVM variation indB per band as a function of the number of spans, withoutlaser linewidth, when employing a Gaussian BS and a 2nd-order super Gaussian BS.

Figure 10 shows that the second and third OFDMbands have similar performance behaviours for both BSimplementations. However, using a Gaussian BS, the firstOFDM band degradation with the number of spans ismuch faster, surpassing the other two bands after 5 spans.As detailed in subsection II-C2, the DP-MZM genera-tion of a SSB signal leaves vestiges of the left sideband.Due to the low selectivity of the Gaussian BS, photo-detection performs the beat between the vestiges of theleft sideband, the desired band and the second band of theright sideband. Together with the aforementioned noiseaccumulation, that also affects the left sideband, this

0 1 2 3 4 5 6 7−40

−38

−36

−34

−32

−30

−28

−26

−24

−22

Number of 40 km spans

EVM

[dB]

1st band

2nd band

3rd band

GaussianBS

2nd−order super

Gaussian BS

Fig. 10: EVM per band as a function of the number ofspans with null laser linewidth, employing a Gaussian anda 2nd-order super Gaussian filter as BSs.

results in a rapid deterioration of the overall performance.Figures 11 and 12 show the log10(BER) results as a func-

tion of the number of spans, for different laser linewidths,when employing a Gaussian BS and a 2nd-order superGaussian BS, respectively. Phase-to-intensity conversionof laser phase noise due to the BS is also confirmed by thelog10(BER) results, since, even in B2B, the MB-OFDMsignal is greatly affected by the increase of laser linewidth.Because of the degradation imposed by the combinationof the BS and phase noise, for laser linewidth higher than1 MHz (just 1 kHz in the 2nd-order super Gaussian BScase) the log10(BER) variation with the number of spansbecome imperceptible.

0 1 2 3 4 5 6 7−6

−5

−4

−3

−2

−1

Number of 40 km spans

log 1

0(BER)

0 Hz

1 kHz

10 kHz

100 kHz

1 MHz

Fig. 11: log10(BER) as a function of the number of spansfor different laser linewidths, employing a Gaussian filteras BS.

The simulation results shown in figure 11 allow de-termining the maximum laser linewidth supported by

PEREIRA: LIMITATIONS OF MB-OFDM METRO NETWORKS BY PHASE-TO-INTENSITY CONVERSION OF LASER PHASE NOISE 9

0 1 2 3 4 5 6 7−6

−5

−4

−3

−2

−1

Number of 40 km spans

log 1

0(BER)

1 kHz

10 kHz

100 kHz

1 MHz

Fig. 12: log10(BER) as a function of the number of spansfor different laser linewidths, employing a 2nd-order superGaussian filter as BS.

the MB-OFDM system and the corresponding maximumreach of the metro ring, when employing a Gaussian BS.According to figure 11, the MB-OFDM signal is able totravel through 5 spans (200 km) before reaching the targetBER. This mark is accomplished for laser linewidths upto 10 kHz. For linewidths of 100 kHz or higher the systemperformance surpasses the target BER even for a singlespan.

Figure 12 shows that, employing a 2nd-order superGaussian BS, with a laser linewidth of 10 kHz the MB-OFDM signal crosses a total of 7 spans (280 km) withoutreaching the target BER. Additionally, figure 12 showsthat this theoretical BS solution supports laser linewidthsup to 100 kHz (for a laser linewidth of 100 kHz results alog10(BER) equal to -2.8).

V. ConclusionsIn this work, the transmission of virtual carrier-assisted

SSB DD MB-OFDM signals along a metro network im-paired by laser phase noise was studied. The impact of thephase-to-intensity noise conversion due to the combinedeffect of finite laser linewidth with fibre CD and BS wasassessed.

The OFDM signal was mathematically defined and theOFDM DFT-based transmission system was described. Avirtual carrier-assisted MB-OFDM system employing SSBsignals generated by an optical transmitter using a DP-MZM and employing an optical receiver supported by PINphoto-detection was presented. Phase noise effects on DD-OFDM systems were identified as power degradation, PRTand ICI.

Results were presented for the implementation of a nu-merical simulator based on the MB-OFDM metro networksystem modelled. The MB-OFDM system performancewas evaluated by two different figures of merit: the re-quired OSNR to achieve a BER of 10−3 and the BER

value. It was shown that due to phase-to-intensity noiseconversion a OSNR increase of 1 dB is necessary for a 5MHz laser linewidth after a single SSMF span of 500 km.Considering noise amplification and accumulation, it wasshown that the MB-OFDM signal can reach 5 SSMF spanswith 40 km before reaching a BER = 10−3, when em-ploying a Gaussian BS, while employing a 2nd-order superGaussian BS the MB-OFDM signal can travel through atleast 7 spans. These results were possible for a maximumlaser linewidth of 10 kHz, employing a Gaussian BS, whilethe 2nd-order super Gaussian BS solution allowed laserlinewidths up to 100 kHz.

It can be concluded that for distances within the metronetwork range, the impact of phase-to-intensity conversionof laser phase noise associated with fibre CD in the systemperformance is small, at least for laser linewidths up to 5MHz. However, in this work, it was found that phase-to-intensity noise conversion associated with the BS has highdegrading impact on the system performance. The impactis such that, unless laser linewidths are reduced to valuesunder 100 kHz, even highly selective filters as the 2nd-order super Gaussian are unsuitable BS solutions for DDMB-OFDM metro networks.

References[1] A. Saleh and J. Simmons, “Architectural principles of optical re-

gional and metropolitan access networks”, Journal of LightwaveTechnology, vol. 17, no. 12, pp. 1-18, 1999.

[2] N. Madamopoulos, D. Friedman, I. Tomkos, and A. Boskovic,“Study of the performance of a transparent and reconfigurablemetropolitan area network”, Journal of Lightwave Technology,vol. 20, no. 6, pp. 937-945, 2002.

[3] I. Tomkos, “Metropolitan area optical networks”, IEEE Circuits& Devices Magazine, vol. 19 , no. 4, pp. 24-30, 2003.

[4] T. Alves, L. Mendes and A. Cartaxo, “High granularity multi-band OFDM virtual carrier-assisted direct-detection metro net-works”, Journal of Lightwave Technology, vol. 33, no. 1, pp. 42-54, 2015.

[5] W. Wei, D. Xu, D. Qian, P. Ji, T. Wang, and C. Qiao,“Demonstration of an optical OFDMA metro ring network withdynamic sub-carrier allocation”, Proceedings of Optical FiberCommunication Conference, paper NTuA3, San Diego, USA,2009.

[6] R. Davey, D. Grossman, M. Wiech, D. Payne, D. Nesset, A.Kelly, A. Rafael, S. Appathurai, and S. Yang,“Long-reach pas-sive optical networks”, Journal of Lightwave Technology, vol.27, no. 3, pp. 273-291, 2009.

[7] N. Cvijetic, M. Huang, E. Ip, Y. Huang, D. Qian, and T.Wang, “1.2 Tb/s symmetric WDM-OFDMA-PON over 90 kmstraight SSMF and 1:32 passive split with digitally-selectiveONUs and coherent receiver OLT”, Proceedings of Optical FiberCommunication Conference, paper PDPD7, Los Angeles, USA,2011.

[8] L. Mendes, Optimization of the degradation induced by theMZM and PIN nonlinear effects in MB-OFDM metropolitannetworks, M. Sc. Dissertation, Departamento de EngenhariaElectrotecnica e de Computadores, Instituto Superior Tecnico,Lisbon, Portugal, April 2014.

[9] B. Schrenk, J. Lazaro, D. Klonidis, F. Bonada, F. Saliou, V.Polo, E. Lopez, Q. Le, P. Chanclou, L. Costa, A. Teixeira, S.Chatzi, I. Tomkos, G. Beleffi, D. Leino, R. Soila, S. Spirou,G. Valicourt, R. Brenot, C. Kazmierski, and J. Prat, “Demon-stration of a remotely dual-pumped long-reach PON for flexibledeployment”, Journal of Lightwave Technology, vol. 30, no. 7,pp. 953-961, 2012.

[10] J. Homa and K. Bala, “ROADM architectures and their en-abling WSS technology”, IEEE Communications Magazine, vol.46, no. 7, pp. 150-154, 2008.

10 NOVEMBER 2015

[11] S. Blouza, E. Le Rouzic, N. Brochier, B. Cousin, E. Pinceminand J. Karaki, “Multi-band OFDM for optical networking”,IEEE EUROCON 2011 - International Conference on Com-puter as a Tool, pp. 1-4, Lisbon, Portugal, 2011.

[12] E. Basch, R. Egorov and S. Gringeri, “Architectural tradeoffs forreconfigurable dense wavelength-division multiplexing systems”,IEEE Journal of Selected Topics in Quantum Electronics, vol.12, no. 4, pp. 615-626, 2006.

[13] J. Armstrong, “OFDM for optical communications”, Journal ofLightwave Technology, vol. 27, no. 3, pp. 189-204, 2009.

[14] A. Lowery, L. Du, and J. Armstrong, “Orthogonal frequencydivision multiplexing for adaptive dispersion compensation inlong haul WDM systems”, Proceedings of Optical Fiber Com-munication Conference, paper PDP39, Anaheim, USA, 2006.

[15] S. Jansen, I. Morita, N. Takeda and H. Tanaka, “20-Gb/sOFDM transmission over 4,160-km SSMF enabled by RF-pilot tone phase noise compensation”, Proceedings of OpticalFiber Communication Conference, paper PDP15, Anaheim,CA, USA, 2007.

[16] S. Jansen, I. Morita, T. Schenk, N. Takeda and H. Tanaka,“Coherent optical 25.8-Gb/s OFDM transmission over 4160-kmSSMF”, Journal of Lightwave Technology, vol. 26, pp. 6–15,2008.

[17] N. Cvijetic, “OFDM for next generation optical access net-works”, Journal of Lightwave Technology, vol. 30, no. 4, pp.384-398, 2012.

[18] W. Shieh, “OFDM for flexible high-speed optical networks”,Journal of Lightwave Technology, vol. 29, no. 10, pp. 1560-1577,2011.

[19] S. Kim, K. Seo and J. Lee, “Spectral efficiencies of channel-interleaved bidirectional and unidirectional ultradense WDM formetro applications”, Journal of Lightwave Technology, vol. 30,no. 2, pp. 229-233, 2012.

[20] W. Peng, I. Morita, H. Takahashi and T. Tsuritani, “Transmis-sion of high speed (>100 Gb/s) direct-detection optical OFDMsuperchannel”, Journal of Lightwave Technology, vol. 30, no. 12,pp. 2025-2034, 2012.

[21] T. Alves, A. Alberto, and A. Cartaxo. “Direct-Detection Multi-Band OFDM Metro Networks Employing Virtual Carriers andLow Receiver Bandwidth”, Proceedings of Optical Fiber Com-munication Conference, paper Tu3G.5, San Francisco, USA,2014.

[22] B. Schmidt, A. Lowery, and L. Du,“Low sample rate transmit-ter for direct-detection optical OFDM”, Proceedings of OpticalFiber Communication Conference, paper OWM4, San Diego,USA, 2009.

[23] A. Armada, “Understanding the effects of phase noise in or-thogonal frequency division multiplexing”, IEEE Transactionson Broadcasting, vol. 47, no. 2, pp. 153-159, 2001.

[24] L. Tomba, “On the effect of Wiener phase noise in OFDMsystems”, IEEE Transactions on Communications, vol. 46, no.5, pp. 580-583, 1998.

[25] G. Qi, J. Yao, J. Seregelyi, S. Paquet, C. Belisle, X. Zhang, K.Wu and R. Kashyap, “Phase noise analysis of optically gener-ated millimeter-wave signals with external optical modulationtechniques”, Journal of Lightwave Technology, vol. 24, no. 12,pp. 4861-4875, 2006.

[26] X. Yi, W. Shieh and Y. Ma, “Phase noise effects on highspectral efficiency coherent optical OFDM systems”, Journal ofLightwave Technology, Vol. 26, No. 10, pp. 1309-1316, 2008.

[27] M. El-Tanany, Y. Wu and L. Hazy, “Analytical modeling andsimulation of phase noise interference in OFDM-based televi-sion terrestrial broadcasting systems”, IEEE Transactions onBroadcasting, vol. 47, no. 1, pp. 20-31, 2001.

[28] A. Armada and M. Calvo, “Phase noise and sub-carrier spacingeffects on the performance of an OFDM communication sys-tem”, IEEE Communication Letters, vol. 2, pp. 11-13, 1998.

[29] J. Sttot, “The effect of phase noise in COFDM”, EBU TechnicalReview, 1998.

[30] W. Peng, “Analysis of laser phase noise effect in direct-detectionoptical OFDM transmission”, Journal of Lightwave Technology,vol. 28, no. 17, pp. 2526-2536, 2010.

[31] W. Shieh and I. Djordjevic, Orthogonal Frequency-DivisionMultiplexing for Optical Communications, Elsevier AcademicPress, New York, USA, 2010.

[32] R. Peled and A. Ruiz, “Frequency domain data transmission us-ing reduced computational complexity algorithms”, ProceedingsIEEE ICASSP 80, vol. III, pp. 964-967, Denver, CO, USA, 1980.

[33] T. Alves and A. Cartaxo, “Performance degradation due toOFDM-UWB radio signal transmission along dispersive single-mode fiber”, IEEE Photonics Technology Letters, vol. 21, no. 3,pp. 158-160, 2009.

[34] S. Blouza, E. Le Rouzic, N. Brochier and B. Cousing, “Opticalmulti-band OFDM switching technology”, 17th European Con-ference on Networks and Optical Communications, Vilanova ila Geltru, Spain, 2012.

[35] Z. Li, X. Xiao, T. Gui, Q. Yang, R. Hu, Z. He, M. Luo, C. Li, X.Zhang, D. Xue, S. You and S. Yu, “432-Gb/s direct-detectionoptical OFDM superchannel transmission over 3040-km SSMF”,IEEE Photonics Technology Letters, vol. 25, no. 15, pp. 1524-1526, 2013.

[36] W. Peng, X. Wu, V. Arbab, K. Feng, B. Shamee, L. Christen, J.Yang, A. Willner, and S. Chi, “Theoretical and experimental in-vestigations of direct-detected RF-tone-assisted optical OFDMsystems”, Journal of Lightwave Technology, vol. 27, no. 10, pp.1332-1339, 2012.

[37] W. Rosenkranz, A. Ali, and J. Leibrich,“Orthogonal frequencydivision multiplexing (OFDM) in optical communications withdirect detection for metro networks”, International Conferenceon Transparent Optical Networks, paper Mo.B2.3, Azores, Por-tugal, 2011.

[38] W. Peng, B. Zhang, K. Feng, X. Wu, A. Willner and SienChi ”Spectrally efficient direct-detection OFDM transmissionincorporating a tunable frequency gap and an iterative detectiontechniques”, Journal of Lightwave Technology, vol. 27, no. 24,pp. 5723-5735, 2009.

[39] A. Li, D. Che, X. Chen, Q. Hu, Y. Wang and W. Shieh,“61 Gbits/s direct-detection optical OFDM based on blockwisesignal phase switching with signal-to signal beat noise cancella-tion”, Optics Letters, Vol. 38, no. 14, pp. 2614-2616, 2013.

[40] D. Fonseca, A. Cartaxo and P. Monteiro, “Modelling and exper-imental validation of an x-cut four phase modulators structure”,Proceedings Optoelectronics, vol. 153, no. 4, pp. 145-151, 2006.

[41] T. Alves and A. Cartaxo, “Performance comparison of powerfading mitigation techniques in multiband OFDM-UWB signalstransmission along LR-PONs”, Chinese Optics Letters, vol. 11,no. 8, pp. 080602-1 - 080602-6, 2013.

[42] S. Nezamalhosseini, L.Chen, Q. Zhuge, M. Malekiha, F. Mar-vasti and D. Plant, “Theoretical and experimental investigationof direct detection optical OFDM transmission using beat in-terference cancellation receiver”, Optics Express, vol. 21, no. 13,pp. 15237-15246, 2013.

[43] A. Cartaxo, Transmissao por Fibra Optica, Departamento deEngenharia Electrotecnica e de Computadores, Instituto Supe-rior Tecnico, Lisbon, Portugal, 2005.

[44] A. Cartaxo and T. Alves, “Theoretical and experimental per-formance evaluation methods for DD-OFDM systems with op-tical amplification”, Journal of Microwaves, Optoelectronics andElectromagnetic Applications, vol. 10, no. 1, pp. 82-94, 2011.